Optical binary de Bruijn networks for massively parallel computing: design methodology and feasibility study.

Area of Science:

• Computer Science
• Electrical Engineering
• Optical Computing

Background:

• Interconnection network structure significantly impacts parallel computer cost and performance.
• Hypercubes are popular but suffer from increasing node complexity with network size.
• Binary de Bruijn (BdB) networks offer advantages like constant node degree and fewer links, but face VLSI implementation challenges.

Purpose of the Study:

• To propose an implementation methodology for optical binary de Bruijn (BdB) networks.
• To address the challenges of implementing asymmetrical and globally connected BdB networks using optical technology.
• To demonstrate the feasibility of the proposed optical implementation methodology.

Main Methods:

• Developing a novel implementation methodology for optical BdB networks.
• Partitioning the network into primitive, efficiently implementable operations.
• Proposing a specific optical implementation of the BdB network based on the methodology.

Main Results:

• The proposed methodology enables efficient implementation of optical BdB networks.
• The design is shown to be feasible through a concrete optical implementation proposal.
• Optical implementation overcomes limitations of traditional VLSI for BdB networks.

Conclusions:

• Optical technology is well-suited for implementing binary de Bruijn (BdB) networks due to its properties.
• The presented methodology provides a viable approach for constructing large-scale parallel systems using optical BdB networks.
• This work facilitates the realization of high-performance, scalable parallel computers through optical interconnects.